The browser you are using is not supported by this website. All versions of Internet Explorer are no longer supported, either by us or Microsoft (read more here: https://www.microsoft.com/en-us/microsoft-365/windows/end-of-ie-support).

Please use a modern browser to fully experience our website, such as the newest versions of Edge, Chrome, Firefox or Safari etc.

Estimating soil salinity over a shallow saline water table in semiarid Tunisia

Author

Summary, in English

Rapid and reliable observations of soil electrical conductivity are essential in order to maintain sustainable irrigated agriculture. Direct measurement of the electrical conductivity of saturated soil paste (ECe), however, is tedious and
time consuming. Therefore, there are needs to find efficient indirect methods to predict the soil salinity from other readily available observations. In this paper we explore the application of multiple linear regression (MLR) and artificial neural networks (ANN) to predict ECe variation from easily measured soil and groundwater properties under highly complex and heterogeneous field conditions in semiarid Tunisia. We compare two methods for dividing the data set into training and validation sub-sets; a statistical (SD) and a random data set division (RD), and their effect on model performance. The input variables were chosen from the plot coordinates, groundwater table properties (depth, electrical conductivity, piezometric level), and soil particle size at 5 depths. The results obtained with ANN and MLR indicate that the statistical properties of data in the training and validation sets need to be taken into account to ensure that optimal model performance is achieved. The SD can be considered as a solution to resolve the problem of over-fitting a model when using ANN. For the SD, the determination coefficient (R 2 ) when using an ANN model varied from 0.85 to 0.88 and the root mean square error from 1.23 to 1.80 dS m -1 . Because of the complexity of the field soil salinity process and the spatial variability of the data, this clearly indicates the potential to use ANN models to predict ECe.

Topic

  • Water Engineering
  • Other Social Sciences

Status

Published

ISBN/ISSN/Other

  • ISSN: 1874-3781